An Improved YOLO Algorithm for Fast and Accurate Underwater Object Detection

Abstract

Due to the abundant natural resources of the underwater world, autonomous exploration using underwater robots has become an effective technological tool in recent years. Real-time object detection is critical when employing robots for independent underwater exploration. However, when a robot detects underwater, its computing power is usually limited, which makes it challenging to detect objects effectively. To solve this problem, this study presents a novel algorithm for underwater object detection based on YOLOv4-tiny to achieve better performance with less computational cost. First, a symmetrical bottleneck-type structure is introduced into the YOLOv4-tiny’s backbone network based on dilated convolution and 1 × 1 convolution. It captures contextual information in feature maps with reasonable computational cost and improves the mAP score by 8.74% compared to YOLOv4-tiny. Second, inspired by the convolutional block attention module, a symmetric FPN-Attention module is constructed by integrating the channel-attention module and the spatial-attention module. Features extracted by the backbone network can be fused more efficiently by the symmetric FPN-Attention module, achieving a performance improvement of 8.75% as measured by mAP score compared to YOLOv4-tiny. Finally, this work proposed the YOLO-UOD for underwater object detection through the fusion of the YOLOv4-tiny structure, symmetric FPN-Attention module, symmetric bottleneck-type dilated convolutional layers, and label smoothing training strategy. It can efficiently detect underwater objects in an embedded system environment with limited computing power. Experiments show that the proposed YOLO-UOD outperforms the baseline model on the Brackish underwater dataset, with a detection mAP of 87.88%, 10.5% higher than that of YOLOv4-tiny’s 77.38%, and the detection result exceeds YOLOv5s’s 83.05% and YOLOv5m’s 84.34%. YOLO-UOD is deployed on the embedded system Jetson Nano 2 GB with a detection speed of 9.24 FPS, which shows that it can detect effectively in scenarios with limited computing power.

Estimating catch rates in real time: Development of a deep learning based Nephrops (Nephrops norvegicus) counter for demersal trawl fisheries

“…It can predict the bounding box coordinates and the corresponding confidence scores with one single network. There are numerous YOLO versions dedicated to operating on underwater images for detection of various objects such as starfish, shrimp, crab, scallop, and waterweed (Liu et al, 2020;Zhao et al, 2022). Among these models, the recently proposed model, YOLOfish was designed for fish detection and is reported to be performing close to YOLOv4 model on two different public datasets (Muksit et al, 2022).…”

YOLOv5 based detector for eight different urine particles components on single board computer

“…Applying CNN to entire regions is computationally very expensive. To overcome this problem, one of the fastest and highest‐performing architectures widely used in object detection problems is preferred the YOLO family 19–21 …”

“…To overcome this problem, one of the fastest and highest-performing architectures widely used in object detection problems is preferred the YOLO family. [19][20][21] The YOLOv5-based urine analysis system developed in this paper to identify and determine the number of erythrocytes, leukocytes, yeast, epithelium, bacteria, crystals, cylinders and other particles that may be present in the urine is as in Figure 4. The image taken from the microscope is given as input to the YOLOv5 architecture realized in the Python programming language.…”

UTD-YOLO: underwater trash detection model based on improved YOLOv5

“…By integrating a transformer encoder and a coordinate attention module to boost performance, Liu et al 9 suggested an underwater target identification approach for mobile deployment. Zhao et al 10 fused YOLOv4-tiny structure, symmetric FPN-Attention module, symmetric bottleneck-type expanding convolutional layer, and label smoothing training strategy to improve YOLOv4. A lightweight underwater target identification technique was proposed by Zhang et al, 11 which is based on the MobileNetv2, YOLOv4 algorithm, and attention feature fusion.…”

“…suggested an underwater target identification approach for mobile deployment. Zhao et al 10 . fused YOLOv4-tiny structure, symmetric FPN-Attention module, symmetric bottleneck-type expanding convolutional layer, and label smoothing training strategy to improve YOLOv4.…”